Anthropic Releases Opus 5 — Nearly on Par with Fable 5, but Half the Price
Translated by a model; the original is in Russian.
Anthropic has released Claude Opus 5 — a new model the company describes as “approaching the capabilities of Claude Fable 5 in many areas” and noticeably stronger at complex programming tasks. Fable 5 is a public Mythos-class model that previously drew the ire of the U.S. authorities: together with Mythos 5, it was taken offline for several weeks and then brought back with reinforced cyberdefense mechanisms.
The release came weeks after yet another standoff between Anthropic and the U.S. government, and a few days after an industry-wide security incident at OpenAI. The Fable 5 saga and the negotiations that followed, according to The Verge, ushered in a new era of AI regulation under the Trump administration: OpenAI's GPT-5.6, which came out soon afterward, went through a roughly two-week “probation period,” being available at first only to state-approved organizations.
Asked whether Anthropic had cleared Opus 5 with the authorities before its public launch, company spokesperson Danielle Guillery said Anthropic continues to work with “government partners” on independent testing of its models, including Opus 5.
Betting on Enterprise Customers and Safety
Opus 5 is being pitched first and foremost to businesses: the company calls it its best model for knowledge work and tasks in biology, and “the perfect model for everyday use.” The most ambitious projects and long-running agentic scenarios still belong to Fable 5. At the same time, Anthropic stresses that Opus 5 is more secure than its predecessor Opus 4.8, is the “most aligned” of the Opus line, and is the least susceptible to attempts to trick it into harmful use.
With this release the company appears to be carefully addressing recent complaints as well — disputes over allegedly hidden usage limits (which led to at least one lawsuit) and fears that AI labs' costs are being passed on to customers. The price per million tokens is unchanged — $5 for input and $25 for output, which, as Anthropic notes, is half the price of Fable 5 and slightly cheaper than OpenAI's GPT-5.6.
Users also have access to Fast mode (launched as a research preview): it delivers twice the speed for twice the price. In addition, you can enable an automatic fallback to a lower-tier model in cases where its safety mechanisms force Opus 5 to decline a request.
Source: The Verge
Comments
Sign in, to comment.